sslf 0.2.0

A simple spectral line finder

sslf is designed to be a dead-simple, effective and useful spectral line finder for 1D data. It utilises the continuous wavelet transform from scipy, which is a productive way to find even weak spectral lines.

Because there are many definitions for a spectral line, this package can not conceivably work in all cases. Here, I have designed sslf for Gaussian-profiled spectral lines. A big part of my work focuses on recovering weak signals in the noise, so sslf aims to identify lines, flag them, and subtract the non-zero bandpass around them.

Usage

With a 1D spectrum (spectral_data) and a range of scales to test (e.g. 1 through 20), statistically significant spectral peaks can be found with:

import numpy as np
from sslf.sslf import Spectrum

s = Spectrum(spectral_data)
s.find_cwt_peaks(scales=np.arange(1, 20), snr=6.5)
s.subtract_bandpass()

The flattened spectrum is then contained in s.modified, and peak locations at s.channel_peaks. Other parameters related to the peaks are s.peak_snrs, s.vel_peaks and s.channel_edges.

Note that the snr parameter specified to find_cwt_peaks actually refers to the signal-to-noise ratio in the wavelet domain, not the input-signal domain; for this reason, if you wanted all signals above 5-sigma, you would have to put in a little more work to determine what your spectrum’s noise RMS is and how it maps to the wavelet RMS.

Also note that channel_edges can be quite inaccurate; by default, this variable is populated by the wavelet scale that found the spectral line. The wavelet need not match very well; it only needs to be significant enough to be picked up by sslf. The Spectrum object has an optional refine_line_widths method which (hopefully) does a better job of finding the channel extent of all detected spectral lines. Read the documentation of sslf functions for more information.

find_cwt_peaks can optionally take the wavelet to be used in the wavelet transformation (Ricker wavelet by default). subtract_bandpass is a little harder to explain; more complicated examples of usage can be found in the notebooks directory. A reasonably complex example of sslf usage is here.

Contributions and Suggestions

I created this software because I could not find anything in the community that functioned similarly. So, I would welcome contributions to make this software more generally flexible, to be suitable for a wide audience. It’s really simple, just look at the code!

Issues? Bugs? Inconsistencies?

Absolutely! Please raise a GitLab issue if you find something odd, with your inputs and expected outputs, and I’ll attempt to fix it ASAP.

Installation

pip install sslf

or

• clone this repo (git clone https://gitlab.com/cjordan/sslf.git)

• within the directory, run: python setup.py install --optimize=1

Dependencies

• python 2.7.x or 3.x

• numpy 1.8+

• scipy

• future